Method, system and equipment for predicting micro-water content in transformer

Abstract

The invention belongs to the related technical field of transformer monitoring, and provides a method, a system and equipment for predicting the micro-water content in a transformer in order to solve the problems of poor real-time performance and inaccuracy of existing transformer micro-water prediction. The method comprises the following steps: acquiring an association relationship between different insulation weak parts and the micro-water content of an oil taking port under different operation parameters and / or external environment parameters of a transformer, further constructing a plurality of training sample sets, and respectively constructing a dynamic deduction model of the micro-water content of the transformer on the basis of the difference of micro-water diffusion characteristics and temperature / electric field influence degrees of the weak parts, so as to obtain a dynamic deduction model of the micro-water content of the transformer. The averaging processing of a single model on the characteristics of different parts is avoided, the specific correlation between each weak part and the micro-water content of the oil taking port can be more accurately captured, and the accuracy of a deduction result is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field related to transformer monitoring, and in particular relates to a method, system and equipment for predicting the trace water content inside a transformer. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Moisture is widely present in transformer insulating oil and paperboard due to moisture intrusion and insulation aging, making it one of the most common insulation defects in oil-paper insulation. Increased moisture content in oil-paper insulation not only accelerates the aging process but also causes electric field distortion, inducing and accelerating partial discharge. In engineering applications, offline measurement is often used. During power outages or sampling and analysis, samples are taken from the transformer oil tapping port. Karl Fischer titration and the oil-paper equilibrium curve are used to estimate the moisture content in the solid insulation. Offline measurement only involves sampling during power outages and cannot provide real-time monitoring of the transformer's solid insulation. Offline measurement is indirect, relying on an ideal oil-paper equilibrium curve to estimate the moisture content in the solid insulation paperboard. However, in actual operation, the uneven temperature and moisture distribution make it difficult to achieve an ideal equilibrium state, resulting in poor calculation accuracy. Sampling only from the oil tapping port only reflects the trace moisture content of the insulating oil at that tapping port, but does not represent the moisture distribution of the insulating oil throughout the transformer, making it difficult to determine the location of weak insulation within the transformer.

[0004] Furthermore, online monitoring methods exist only as auxiliary detection methods, requiring the installation of sensors such as distributed fiber optic sensors and miniature humidity sensors. However, during operation, transformers are subject to the coupling of multiple physical fields, including electric, temperature, and fluid fields. The distribution of these fields is extremely uneven, and the high voltage levels and harmonic content place higher demands on the sensor's insulation level, high-temperature resistance, and corrosion resistance, which in turn interferes with the sensor detection process and data accuracy. Furthermore, the compact interior space and numerous insulation structures of transformers make it difficult to install physical miniaturized sensors in areas with weak insulation, thus failing to address the practical problem of detecting the distribution of moisture content at different locations in the transformer.

[0005] Therefore, how to accurately capture the dynamic distribution of transformer micro-water without relying on the large-scale deployment of physical sensors that are susceptible to interference from multiple physical fields in a complex internal environment, while avoiding the indirect estimation errors and sampling limitations based on the oil-paper balance assumption in offline detection, is an urgent problem that needs to be solved. Summary of the Invention

[0006] To overcome the deficiencies of the above-mentioned prior art, the present invention provides a method, system, and equipment for predicting the micro-water content inside a transformer, constructs multiple dynamic deduction models for the micro-water content of a transformer, accurately captures the specific correlation between the weak points of the transformer and the micro-water content of the oil tap, and improves the accuracy of the results.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for predicting the moisture content inside a transformer, comprising: By simulating the transformer multi-physics field coupling model, the correlation between different insulation weaknesses and the water content at the oil tapping port under different transformer operating parameters and / or external environmental parameters is obtained, and multiple training data sets are constructed; The regression decision tree model is trained based on multiple training sample sets to obtain a dynamic deduction model of transformer moisture content corresponding to different insulation weaknesses. Based on the obtained micro-water content at the oil port of the transformer to be predicted, the micro-water content of different weak insulation parts inside the transformer is obtained by using the dynamic deduction model of transformer micro-water content corresponding to different weak insulation parts.

[0008] In a second aspect, the present invention provides a system for predicting the moisture content inside a transformer, comprising: The construction module is configured to: simulate the transformer multi-physics field coupling model to obtain the correlation between different insulation weaknesses and the water content at the oil tap under different transformer operating parameters and / or external environmental parameters, and construct multiple training data sets; The training module is configured to: train a regression decision tree model based on multiple training sample sets to obtain a dynamic deduction model of transformer micro-water content corresponding to different insulation weaknesses; The prediction module is configured to: based on the obtained micro-water content at the oil port of the transformer to be predicted, use the transformer micro-water content dynamic deduction model corresponding to different insulation weak points to obtain the micro-water content of different insulation weak points inside the transformer.

[0009] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0010] One or more of the above technical solutions have the following beneficial effects: In the present invention, simulation is performed by using a transformer multi-physics field coupling model to obtain the correlation between different insulation weaknesses and the micro-water content at the oil tapping port under different transformer operating parameters and / or external environmental parameters, and then multiple training sample sets are constructed. Based on the differences in micro-water diffusion characteristics and the degree of influence of temperature / electric field at each weak point, dynamic deduction models of transformer micro-water content are constructed respectively, avoiding the averaging of characteristics of different parts by a single model, and being able to more accurately capture the specific correlation between each weak point and the micro-water content at the oil tapping port, thereby improving the accuracy of the deduction results.

[0011] In this invention, the dynamic deduction model based on a regression decision tree can be combined with an incremental update mechanism. When new operating data arrives, such as new data on the moisture content at the oil tap, the node weights are adjusted to adapt to the data changes. This feature enables the model to track the dynamic changes in moisture content within the transformer over the long term, avoiding the loss of deduction accuracy caused by equipment aging and evolving operating conditions, and ensuring long-term reliability.

[0012] In the present invention, the tree complexity regularization term penalizes the number of nodes in the regression decision tree to avoid the model generating too many branches to fit the training data, thereby preventing over-fitting. The model can more accurately deduce new data while maintaining a good fit to the training data, thereby improving the generalization ability for actual scenarios; the time decay factor reduces the weight of old data, so that the regression decision tree model pays more attention to the changes in the transformer operating status reflected by the new data. This feature is particularly suitable for transformers that operate for a long time, whose internal micro-water distribution will evolve dynamically over time. The time decay factor can help the model adjust its dependence on historical data in a timely manner, ensure that the deduction results are consistent with the current status of the equipment, and improve the accuracy of long-term monitoring.

[0013] In this method, a piecewise linear model is fitted during node splitting and the loss reduction at the split point is calculated. This loss reduction quantifies the effect of different split points, giving node splitting a clear optimization target. Furthermore, the fitting results of the piecewise linear model can intuitively reflect the trend differences in the data on both sides of the split point, enhancing the interpretability of decision tree node splitting and facilitating understanding of the regression decision tree model's processing logic for variables such as transformer moisture content. The clear splitting logic also facilitates incorporating the time decay factor, allowing for more efficient adjustment of node weights during incremental updates and improving the model's adaptability to dynamic data.

[0014] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0016] Figure 1 This is a flow chart of the method for predicting the trace water content inside a transformer in Example 1 of the present invention. DETAILED DESCRIPTION

[0017] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0018] It should be noted that the terms used herein are for describing particular embodiments only and are not intended to limit the exemplary embodiments according to the present invention.

[0019] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0020] Example 1 This embodiment discloses a method for predicting the trace moisture content inside a transformer, including: By simulating the transformer multi-physics field coupling model, the correlation between different insulation weaknesses and the water content at the oil tapping port under different transformer operating parameters and / or external environmental parameters is obtained, and multiple training data sets are constructed; The regression decision tree model is trained based on multiple training sample sets to obtain a dynamic deduction model of transformer moisture content corresponding to different insulation weaknesses. Based on the obtained micro-water content at the oil port of the transformer to be predicted, the micro-water content of different weak insulation parts inside the transformer is obtained by using the dynamic deduction model of transformer micro-water content corresponding to different weak insulation parts.

[0021] In this embodiment, simulation is performed through the transformer multi-physics field coupling model to obtain the correlation between different insulation weaknesses and the trace water content at the oil tapping port under different transformer operating parameters and / or external environmental parameters, and then construct multiple training sample sets. Based on the differences in trace water diffusion characteristics and the degree of influence of temperature / electric field at each weak point, dynamic deduction models of transformer trace water content are constructed respectively, avoiding the averaging of characteristics of different parts by a single model, and can more accurately capture the specific correlation between each weak point and the trace water content at the oil tapping port, thereby improving the accuracy of the deduction results.

[0022] The following is a detailed description of the method for predicting the moisture content inside a transformer proposed in this embodiment: Step 1: By simulating the transformer multi-physics field coupling model, the correlation between different insulation weaknesses and the water content at the oil outlet under different transformer operating parameters and / or external environmental parameters is obtained, and multiple training data sets are constructed.

[0023] In this embodiment, the electric field distribution and power loss are solved based on the constructed two-dimensional finite element model of the transformer. The temperature field distribution is calculated using the solved power loss as the heat source, and the material parameters are reacted based on the temperature field distribution to achieve coupling between the electric field and the temperature field. The relationship between the diffusion of micro-water in oil-paper insulation and the temperature field is established based on Fick's second diffusion law. The distribution of micro-water in transformer oil and insulating paperboard is simulated. The material parameters are changed based on the micro-water distribution, and the reaction is exerted on the electric field and the temperature field to achieve coupling between the temperature field and the fluid field.

[0024] First, the geometric size parameters of the transformer are obtained and a transformer simulation grid model is constructed. Specifically, the grid modeling interface is collaboratively called through MATLAB and COMSOL, and the transformer geometric model is established quickly and efficiently using programming languages.

[0025] The MATLAB and COMSOL collaborative mesh modeling interface includes the COMSOL Multiphysics with MATLAB interface for transformer modeling in COMSOL. Using Java to program MATLAB control files, MATLAB controls the transformer geometry, enabling rapid and accurate modeling. Manual meshing, rather than automated meshing, allows for targeted meshing of key areas and ensures convergence and efficient simulation calculations.

[0026] A triangular network is selected for the geometric model network for initial segmentation, and multiple ultra-fine segmentations are performed on key boundaries and areas to complete the overall mesh segmentation of the transformer model.

[0027] Acquire experimental data of transformer-related materials at different temperatures, such as the variation patterns of conductivity and relative dielectric constant with temperature, and use the least squares method to fit the nonlinear variation curve of material parameters with temperature.

[0028] Among them, fitting the nonlinear curve of material parameters changing with temperature includes: obtaining the values ​​of relative dielectric constant and conductivity of transformer insulating oil and insulating paperboard at different temperatures, and obtaining the functional relationship of relative dielectric constant γ and conductivity σ changing with temperature T by least squares fitting, where c represents the micro-water content in the transformer oil-paper insulation, as shown in Table 1.

[0029] Table 1:

[0030] The electromagnetic field dynamic parameters are set for the two-dimensional finite element model of the transformer, and the electromagnetic analysis module is used to solve the power loss and electric field distribution, and then the temperature field distribution is obtained. The temperature field distribution determines the material parameters at each part.

[0031] Fick's second diffusion law is used to establish a model of the influence of micro-water diffusion and temperature field in oil-paper insulation. The fluid field adopts the dilute species transfer module. Finally, the micro-water content at each position in the fluid field directly affects the material parameters, realizing the coupling of electric field, temperature field and fluid field.

[0032] The power loss includes core loss and winding loss. Core loss includes hysteresis loss, eddy current loss and abnormal loss, while winding loss includes ohmic loss and eddy current loss.

[0033] The core loss expression is shown in formula (1): (1) in, is the sum of hysteresis loss and abnormal loss, is the eddy current loss. B is the function of magnetic flux density, s is the thickness ratio of silicon steel lamination, f Indicates frequency. Coefficient 、 , α and β are obtained by fitting the input loss curve data.

[0034] When high-order harmonics are present, the eddy current loss under distorted magnetic flux is derived as shown in formula (2): (2) in, is the electrical conductivity of the silicon steel sheet, is the thickness of a single silicon steel sheet, NH is the highest harmonic order, For the The amplitude of the subharmonic.

[0035] The calculation formula for the ohmic loss generated by the winding under harmonic conditions is shown in formula (3): (3) in, It is Subharmonic current and the fundamental current RMS The ratio, is the ohmic loss of the winding under fundamental wave conditions.

[0036] The calculation formula of eddy current loss of winding under harmonic conditions is shown in formula (4): (4) in, is the winding eddy current loss at the fundamental frequency, It is Subharmonic current frequency and fundamental current frequency The ratio.

[0037] The mathematical expression of Fick's second law of diffusion is: (5) Where, is the water concentration in the insulating paper (%), is the diffusion coefficient of water in the insulating paper (m 2 / s), T represents temperature, Represents the gradient.

[0038] The empirical calculation formula of the diffusion coefficient D is shown in formula (6): (6) Where T0 is the reference temperature; Ea is the activation energy (K) during the diffusion process; D0 is a pre-exponential factor, unit m 2 / s; k is a dimensionless parameter.

[0039] The calculation model of water diffusion in oil-paper insulation based on Fick's second diffusion law and the empirical calculation formula of the water diffusion coefficient D are input into the dilute species transfer physics field to build a thermal-mass coupling interface, thereby achieving the connection between the electrothermal coupling model and the related interfaces of the thermal-mass coupling model.

[0040] By establishing the relationship between the temperature field and the diffusion of micro-water, a coupled model of heat and mass transfer inside the mutual inductor was established, realizing the construction of the electric-heat-flow multi-physics field coupled model. Finally, the mesh was divided using COMSOL free triangle mesh, and the volume heat source was applied using the Joule heat obtained from the above calculation. The boundary of the solid heat transfer module was set to convective heat flux 28 In the porous rare earth species transfer module, the porosity of the insulating paper is set to 60% at the oil-paper insulation boundary, and the transient field is used for simulation.

[0041] In this embodiment, based on the constructed transformer electric-thermal-moisture multi-physics field coupling model, the distribution of micro-water content inside the converter transformer under normal operating conditions is obtained, and based on actual engineering experience, multiple micro-water content concentration points are selected as typical insulation weak points.

[0042] By setting different transformer operating parameters and external environmental parameters, the correlation between different insulation weaknesses and the micro-moisture content at the oil tapping port was determined. The converter transformer operating parameters were set, including normal state, damp state, and typical defects. Several typical defects, such as burrs, depressions, and metal discontinuities, were set at the insulation weaknesses. The micro-moisture diffusion characteristics are affected by ambient temperature, humidity, and grid-side and valve-side voltages and currents. By varying these external parameters, the micro-moisture content distribution of the converter transformer was determined under different external parameters.

[0043] In this embodiment, the correlation between different weak insulation points and the trace water content at the oil extraction port is constructed by constructing a data set of different weak insulation points of the transformer and the trace water content at the oil extraction port under different transformer operating parameters and external environmental parameters, and establishing fitting curves of different weak insulation points and the trace water content at the oil extraction port.

[0044] Step 2: Train the regression decision tree model based on multiple training sample sets to obtain a dynamic deduction model of transformer micro-water content corresponding to different insulation weaknesses; based on the obtained micro-water content at the oil port of the transformer to be predicted, use the dynamic deduction model of transformer micro-water content corresponding to different insulation weaknesses to obtain the micro-water content of different insulation weaknesses inside the transformer.

[0045] The training data set established in step 1 is imported into the improved regression decision tree algorithm for training, and then a dynamic deduction model for transformer micro-water content is constructed. Based on the dynamic deduction model for transformer micro-water content, the micro-water content of different insulation weak points inside the converter transformer can be deduced according to the micro-water content at the oil extraction port.

[0046] The present embodiment uses a regression decision tree because the correlation between the oil extraction port and the water content at the weak insulation point is linear, and therefore a regression decision tree is suitable for generating a decision tree model.

[0047] The principle of regression decision tree is: Assume X and Y are input and output variables, respectively, and Y It is a continuous variable. In this embodiment, the concentration of trace water in the insulating oil at the transformer oil inlet is used as the input variable, and the concentration of trace water in the insulating oil paper at the weak insulation point of the transformer is used as the output variable.

[0048] Given a training set , the characteristic parameter set is , is the number of characteristic parameters, The feature space is divided using a heuristic method, and the optimal feature parameter is determined as the split point in each process based on the principle of minimizing the square error.

[0049] To overcome the overfitting problem of traditional algorithms, this embodiment adds a tree complexity regularization term and a time decay factor to the loss function: (7) in, is the number of nodes, which controls the complexity of the regression decision tree model; is the predicted value of the node; and is the time attenuation coefficient; is the regularization strength parameter, represents the loss function, represents the true value of the i-th sample, i.e., the moisture content at the weak insulation point; represents the predicted value of the i-th sample; t represents the time variable. Based on the information loss caused by greedy segmentation, this embodiment introduces piecewise linear regression and spline functions to improve the accuracy of continuous variable processing. When splitting a node, a piecewise linear model is fitted to calculate the loss reduction at the split point s. Linear functions are fitted to the data sets on both sides of the split point. By comparing the loss changes before and after the split, the optimal split point is selected. The loss reduction at the split point s is calculated as: (8) in, and are the left and right fitting values ​​of the piecewise linear function respectively; s is the split point, a node. When calculating the loss function, a split point is selected as S in order to improve the processing accuracy of continuous variables; Refers to the loss function L of the previous step.

[0050] For continuous variables such as temperature T , using cubic spline interpolation: …… (9) in, For the spline knots, it is optimized by minimizing the sum of squared residuals; 、 、 、 、 are the coefficients of the corresponding expansions.

[0051] In this embodiment, the dynamic adaptability is enhanced by combining online learning mechanism to overcome the problem that traditional decision trees are batch learning and cannot adapt to dynamic data streams. Specifically, incremental update rules are introduced to adjust node weights. When new data arrives, only the node weights of the relevant paths are updated: (10) in, is the learning rate, Represents new data, represents the predicted value of new data based on the old weights, The node weight before updating.

[0052] If the new data causes the error of a certain node to continue to increase, triggering a local re-split, you can adjust only the node and its subtree to adjust the local structure.

[0053] In addition, an adaptive learning rate strategy is introduced to dynamically adjust the learning rate according to the error fluctuation amplitude between new data and historical data, and reduce the learning rate when the error fluctuation is large. Avoid overshoot and increase the fluctuation when the fluctuation is small Accelerate convergence and improve the model's ability to track the latest status.

[0054] In this example, the training dataset was imported into the regression decision tree model for training, and the validation dataset was set to 30% of the dataset to verify the output value of the deduction model to improve the training and validation accuracy. The fit of the deduction model was determined by the root mean square error of the deduction model.

[0055] Example 2 The purpose of this embodiment is to provide a system for predicting the moisture content inside a transformer, including: The construction module is configured to: simulate the transformer multi-physics field coupling model to obtain the correlation between different insulation weaknesses and the water content at the oil tap under different transformer operating parameters and / or external environmental parameters, and construct multiple training data sets; The training module is configured to: train a regression decision tree model based on multiple training sample sets to obtain a dynamic deduction model of transformer micro-water content corresponding to different insulation weaknesses; The prediction module is configured to: based on the obtained micro-water content at the oil port of the transformer to be predicted, use the transformer micro-water content dynamic deduction model corresponding to different insulation weak points to obtain the micro-water content of different insulation weak points inside the transformer.

[0056] In further embodiments, there is also provided: An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed by the processor. When the computer instructions are executed by the processor, the method described in Example 1 is performed. For the sake of brevity, no further details are given here.

[0057] It should be understood that in this embodiment, the processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), off-the-shelf field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0058] The memory may include a read-only memory and a random access memory, and provides instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0059] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in embodiment 1 is performed.

[0060] The method in Example 1 can be directly implemented as being executed by a hardware processor, or by a combination of hardware and software modules within the processor. The software module can be located in a storage medium well-established in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not given here.

[0061] Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0062] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A method for predicting the micro-water content inside a transformer, characterized in that: include: By simulating the transformer multi-physics field coupling model, the correlation between different insulation weaknesses and the water content at the oil tapping port under different transformer operating parameters and / or external environmental parameters is obtained, and multiple training data sets are constructed; The regression decision tree model is trained based on multiple training sample sets to obtain a dynamic deduction model of transformer moisture content corresponding to different insulation weaknesses. Based on the obtained micro-water content at the oil port of the transformer to be predicted, the micro-water content of different weak insulation parts inside the transformer is obtained by using the dynamic deduction model of transformer micro-water content corresponding to different weak insulation parts.

2. The method for predicting the micro-water content inside a transformer according to claim 1, wherein: In the training of the regression decision tree model, a segmented linear model is introduced when the node is split. By calculating the loss drop at the split point, linear functions are fitted to the data sets on both sides of the split point respectively. By comparing the loss changes before and after the split, the optimal split point is selected.

3. The method for predicting the micro-water content inside a transformer according to claim 1, wherein: The construction of the transformer multi-physics field coupling model is specifically as follows: Based on the constructed two-dimensional finite element model of the transformer, the electric field distribution and power loss are solved; The temperature field distribution is calculated using the solved power loss as the heat source, and the material parameters are reacted based on the temperature field distribution to achieve the coupling of the electric field and the temperature field; Based on Fick's second diffusion law, the relationship between the diffusion of micro-water in oil-paper insulation and the temperature field is established. The distribution of micro-water in transformer oil and insulating paperboard is simulated. The material parameters are changed based on the micro-water distribution, which reacts to the electric field and temperature field to achieve the coupling of the temperature field and the fluid field.

4. The method for predicting the internal moisture content of a transformer according to claim 1, wherein: By simulating the transformer multi-physics field coupling model, the correlation between different insulation weaknesses and the water content at the oil outlet under different transformer operating parameters and / or external environmental parameters is obtained, and multiple training data sets are constructed, specifically: Based on the transformer multi-physics field coupling model, multiple areas with concentrated trace water content were selected as typical insulation weak points. By setting different transformer operating parameters and / or external environmental parameters, the correlation between different insulation weak points and the trace water content at the oil intake was obtained, and multiple training data sets were constructed.

5. The method for predicting the internal moisture content of a transformer according to claim 4, wherein: The different operating parameters of the transformer specifically include: normal state of the transformer, damp state of the transformer and typical defects of the transformer; the typical defects of the transformer include burr defects, concave defects and metal discontinuity defects.

6. The method for predicting the moisture content inside a transformer according to claim 1, wherein: A tree complexity regularization term and a time decay factor are introduced into the loss function of the regression decision tree model.

7. The method for predicting the moisture content inside a transformer according to claim 1, wherein: Also includes: An incremental update mechanism is introduced to adjust the node weights of the regression decision tree model, so that the node weights are dynamically corrected according to the error of the new data; an adaptive learning rate strategy is introduced to dynamically adjust the learning rate according to the error fluctuation amplitude between the new data and the historical data.

8. The method for predicting the moisture content inside a transformer according to claim 3, wherein: In constructing the transformer multi-physics field coupling model, it also includes: obtaining the values ​​of the relative dielectric constant and conductivity of the transformer insulating oil and insulating paperboard at different temperatures, and obtaining the functional relationship between the relative dielectric constant and conductivity and the temperature by least squares fitting.

9. Transformer internal micro-water content prediction system, characterized by: include: The construction module is configured to: simulate the transformer multi-physics field coupling model to obtain the correlation between different insulation weaknesses and the water content at the oil tap under different transformer operating parameters and / or external environmental parameters, and construct multiple training data sets; The training module is configured to: train a regression decision tree model based on multiple training sample sets to obtain a dynamic deduction model of transformer micro-water content corresponding to different insulation weaknesses; The prediction module is configured to: based on the obtained micro-water content at the oil port of the transformer to be predicted, use the transformer micro-water content dynamic deduction model corresponding to different insulation weak points to obtain the micro-water content of different insulation weak points inside the transformer.

10. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 8 is completed.

Images